Yeah I don't mind explaining in more detail. In fact if that "ultimate smash poll" rears its ugly head somewhere else, the more people who can explain what's wrong with it, the better (especially for us). I am also posting the body of this analysis on reddit to hopefully minimize any skewed impressions it causes there.
Here is the "ultimate smash poll" again:
https://docs.google.com/spreadsheet...pw5m-b4JHJ0_cGRKU15KYlTe1E/edit#gid=729442469. It came from this tumblr:
http://ssb4dojo.com/. This "ultimate smash poll" is not a poll itself, but rather an aggregation of several different past polls that have been done on DLC characters. He obtained his "global" poll results in the following way. He took all Japanese polls he had (just two), combined them directly, and then found the percentage of the total votes that each character got. He did the same thing for all of his "USA/Euro" polls. Then he weighted the Japanese percentages by 1/3, the USA/Euro percentages by 2/3, and then combined these. He then dropped characters that didn't show up on enough polls on the USA/Euro side.
Here is the math for K. Rool as an example. Across all US/Euro polls, he got 2949 votes, which accounted for 6.91% of the votes. Across both Japanese polls, he got 30 votes, which accounted for 3.21% of the votes. 6.91(2/3) + 3.21(1/3) = 5.677%. His number is slightly higher, because the dropped characters inflate everyone else's percentages slightly.
The first and probably biggest problem is that the Japanese results are heavily over-weighted. If you look at the two Japanese polls he included, they are outdated (one is from February, before the announcement of the ballot) and minuscule compared to the others. Thus a tiny number of questionable results are being given a huge influence in the global results he lists.
If you look at his NA/EU results, two polls there are by far the largest and dominate the results, and the data from both is old. One of them in particular (the larger of the two I think) is from before the ballot was announced, and the results look very different from later polls. For example, Shovel Knight doesn't show up at all and Ice Climbers are number 1, even though they aren't even top 5 on any other USA/Euro poll.
In summary, four polls, all of which are outdated, and two of which are tiny, dominate the results. These problems result in the "global" data he compiles being highly distorted, and it shows in the results. Ice Climbers have more than double the votes of Shovel Knight? That seems fishy to say the least.
There are other issues too. Some of the included polls were protected against vote spamming, others were not. There is undoubtedly some degree of cross contamination between polls (in other words if you saw and voted in every poll, then all 8 of your votes counted). Finally there have been more online polls than are included on this list, so it leaves out some potentially significant results.
In conclusion, looking at the individual polls and paying attention to when they are dated will give you some sense of who the popular characters are, but looking at the “global” results as presented there is extremely misleading. Notice that in the more recent NA/EU polls Shantae does very well. This is consistent with the observation that she has risen hugely in popularity and prominence in the months following the opening of the poll. Almost every single video about DLC candidates on YouTube talks about her. We have no good data on Japan whatsoever. The polls he included are interesting to read and not worthless when you take them for what they are, but they don't tell us a lot about what Japanese gamers are voting for right now.
